The bad thing about satellite that makes it unsuitable for many use cases will always remain to be latency, not speed or availability. 1.5 second ping is perfectly usable for web browsing, but it is painful for many other workflows.. And there's no current way with our technology we can "fix" this problem. It boils down to speed of light issues.
I for one am still hoping we'll eventually get quantum internet that goes through the earth rather than around it.
Not really no. LEO is 160-2000km up; it takes light from 1 to 12 milliseconds to do a roundtrip there. This is an order of magnitude faster than it takes your phone or tablet to react to touch input. It'll be more than enough for interactive things.
Add to this that most terrestrial networking is done over fiber optics - the speed of light through the glass is about 2/3 the speed of light in a vacuum. If you're transmitting radio waves into space and between satellites, you'd be as fast as most terrestrial networks so long as the transmitted arc plus twice the altitude was less than 1.5 times the same arc at ground level.
As an aside, the lower speed of light through fiber optic cable is one reason many stock trading platforms use line-of-sight microwave for their primary transmission. For HFT, those uS (and even nS) matter.
I'm using sattelite internet from Cambodia (WiMax). Ping to 8.8.8.8 is around 8ms, and I pay $25 a month for 4mbit. Relability is better than the ADSL or Fiber options over here, due to countries like these being absolutely horrible at wiring cables.
Are you sure it's really going through a satellite? WiMAX isn't per se satellite technology but normal radio with fixed basestations (though satellites are sometimes used to link really remote WiMAX stations)
also I have seen 8.8.8.8 respond from cache kind (don't know how it works) 8ms seems to be impossible from combodia even on wired unless local ISP end is responding to that
Yep, I have a friend who pays $150/month for 50mbit/s satellite with a 20G data cap. He makes heavy use of the unmetered time from 12am-4am though. And yea, I did some look ups before about this issue and it's ridiculous the cap structure some of these providers have, like this bit from Wild Blue:
> After exceeding either of the usage thresholds, which vary by service package, we may significantly slow and/or restrict your service, or certain uses of your service. Your full service will be restored when your data usage in the prior 30 days falls below 70% for both usage thresholds.
So, if you go over, then for the next 30 days you're throttled UNTIL your volume stays under 70% of your cap over a 30 day period... Like, how is it even legal to sell one thing, but punish for using more than 70% of it sometimes
They have sold 100k Porsches on to an information superhighway that has only room for 1000 at a time so they only let each one use it for a relatively short amount of time each month. i.e the do contention management over a month.
Most providers sell consumer broadband access based on headline speed because that's what the fixed line providers do.
When buying "professional" satellite broadband; users can buy un-contented at much lower speeds. A Ford Transit with a dedicated lane perhaps...
Most geostationary orbit satellites are actually closer to 35,000 km above the surface of the planet. As stated by other commenters, LEO, where these microsats would be, is only about 2000 km above the surface. SpaceX wants to put their a mere 1200km[1] up. For most (read non-gamer) use cases, this should be extremely adequate, and it will be at a nice price point if they put up the 4000+ of these that Elon Musk wants to with SpaceX.
Musk is planning to put satellites in LEO, which gives you a roundtrip of 1 to 12 milliseconds (just light; not counting data processing), depending on placement. The 1.5 second ping exist only for current Internet satellites that are parked in GEO.
There's a tradeoff there in the number of satellites needed and the amount of fuel needed to maintain their orbits... the closer to the planet, the more drag. I suspect the larger number of smaller satellites will also produce lower available bandwidth, but I'm not sure how that scales.
I recently saw an article about impulse engines using an arc-welding type approach where just electricity and some sort of metallic rod were needed. If these can be made small enough that could bring down the cost of station keeping dramatically, since presumably you could get a lot more lifetime thrust per kg of mass using that metal, then you would with propellants. (And the electricity can be harvested in orbit.)
"The bad thing about satellite that makes it unsuitable for many use cases will always remain to be latency, not speed or availability. 1.5 second ping is perfectly usable for web browsing, but it is painful for many other workflows."
I have Viasat Exede at my rural home that is unserved by other network products (no mobile signal, no CATV, no DSL).
Ping times are ~700ms and given my high reliance on SSH to do my work I was skeptical ... but it's not that bad.
I can't speak to gaming, etc., but for interactive login and things like skype/vonage it works OK.
Definitely check out Mosh. It's a (roughly! Don't yell at me:) layer ontop of SSH that uses the normal protocol for the initial connection, but after that it uses a UDP based protocol that is heavily tuned for poor network environments.
That, ontop of a whole host of other reasons, is why it's amazing. Absolutely love it, and have been using it every day for over a year now. Despite frequent internet drops (yay, comcast) and laptop closures, i only reconnect to the server when i reboot my laptop (maybe once a month, at best). It's neato!
edit: Note that my "server" work is my IDE (vim), and many panes of tmux, etcetc. Basically all my work. So it's used over 50h a week.
For standard web stuff HTTP2 should make a difference for people with high latencies, possibly some kind of converting proxy might help for legacy sites too.
I don't think anything wireless(besides free space optics) will ever be able to match the bandwidth of fiber. There just isn't enough radio spectrum available. Using currently available COTS equipment you can easily push 100+gb/s over a single fiber. Such capacity isn't feasible even for terrestrial based wireless at this time.
That being said, I hope one of these projects takes off.
Most of the undeveloped world would benefit greatly from even just 5mb/s of internet.
> I don't think anything wireless(besides free space optics) will ever be able to match the bandwidth of fiber.
This isn't really the goal of the project. SpaceX needs to create a com link with Mars. Also, it can cheaply laucnch sattelites so it will be able to connect the world via satelite uplink. The goal isn't to beat wired networks but provide redundancy and in some cases primary connections.
SpaceX is betting on itself achieving reusability which will lower costs from ~$60M to a marginal cost of something like ~$600K. As a byproduct of the cellphone wars nanosats have become insanely cheap due to commoditization and miniturization of hardware. Making this more feasible
The reason they are doing this is to establish communications between earth & mars. Even a 50mbps connection could bring huge swathes of the world online and create a race for infrastructure to compete.
So if we could get 1-5gbps connections terrestriallydue to some market penetration and competition, it would bringlots of people online. The goal of the project is lofty, but would provide a lot of positive things.
Downsides:
* personal satelites and the proliferation of spying.
* space debris could lock us out of orbit.
* space debris/launch failures could cause issues.
* something like radiation or an unexpected externality could make this non-feasible.
* we could literally create skynet.
I am pumped on it though, and those reasons, along with the many others I have assuredly missed make me really really excited to see this happen. We need a secondary network anyway and terrestrial speeds won't improve without competition so I would roll the dice on skynet for that.
I was mostly addressing the comment that the signal could bounce from satellite to satellite 'effectively faster than fiber.'
datenwolf did the math and answered it in greater detail than I.
Didn't Iridium try to do this a while back? Launch a constellation of LEO satellites to provide low latency data and cell communications? Hopefully the reduced launch and operation costs expected will give these current projects more success.
When you say there isn't enough, is that there isn't physically enough spectrum, or is it just that there isn't enough licensable spectrum (or even enough available unlicensed spectrum)?
I am not very knowledgeable about things related to radio spectrums and so please clarify this for me.
What might the upper bound of radio broadcast data transmissions be if the entire radio spectrum could be allocated and given our current radio comms technology?
> When you say there isn't enough, is that there isn't physically enough spectrum, or is it just that there isn't enough licensable spectrum (or even enough available unlicensed spectrum)?
Both. Or rather physical spectrum is a superset of licencable spectrum and already the physical spectrum only has so many quantum numbers to encode information in.
We can pretty much rule out orbital momentum encoding for satellite transmissions; the effects the atmosphere has on polarization are just too strong, for orbital momentum to survive to a degree, that it remains usable for consumer grade equipment.
That leaves as encodable quantum numbers:
- frequency
- amplitude
- polarization
- momentum (i.e. direction of propagation)
The band usable for radio satellite communication, i.e. able to pass through the atmosphere is rather narrow (compared with the whole EM spectrum): About between 100MHz up to 20GHz. That leaves you with a total usable bandwidth of about 19.9GHz.
Multiply that with the symbol density reachable through AM (under good conditions with a low noise floor you can encode about 5 to 6 bits). Polarization gives another bit. Spatial multiplexing gives you log2(n steerable directions) extra bits (usually not more than 4 beams you can control). Divide that by two, because existing transceiver systems (except for some lab systems) can't do full duplex on a given channel.
If you multiply that up this gives you a theoretical limit of about 10TBit/s you can implement for a given satellite location (having several satellites in proximity means they have to share the spatial encoding bits).
On top of that you have the licensing which is going on, so if one of these systems get put into place they will most likely dominate the spectrum so that the others, if deployed, wouldn't even be allowed to transmit.
Good answer by the way. I can second that what datenwolf said is pretty darn accurate, but doesn't account for a lot of the inteference that happens to reduce usability of the specturm.
You also get into beam widths so and antenna. So if you have 700 satellites up in the air given full spectrum that means they'd have a fixed limit of 10TBit/s. That can be divided between their users (but it won't be anything close to 10TBit/s).
So this will really be a factor of how big their beams are, how many people are within their beams, and how much frequency they can legally use. While this is really cool and possible, I doubt this will open up a whole new internet in it's first pass.
There is also the issue with, how are they going to get the signal to it's destination? They can't beam it all the way around the world for you, and from what I understand at least OneWeb isn't going to communicate from satellite to satellite. So they'll have to have a ton of ground stations to receive your link and then send it over fiber, or transmit up to GEO, then down again. OR they'll have to hop up and down between nodes to it's destination, in a similar fashion to TOR. I don't know how they plan to solve this or if SpaceX has a better solution.
On a slight tangent Shannon had a lot to say about signal to noise ratios / bit error rates and you're not going to like the theoretical minimum transmitter power required to push 10 TB over 20 GHz of BW at a reasonable bit error rate into orbit. Now figure maybe twice that for practical implementation at that huge scale, and twice that again because ultra wide band amplifiers are not known for cheapness or efficiency... This theoretical satellite phone from 2050 is going to get somewhat warm in your hand...
Oh, I'm fully aware of the information theoretical limit. I just made an back-of-the-envelope estimation on the absolute theoretical upper bound, assuming a system operating at 0K and zero noise floor. My intention was to show, how small the available bandwidth is already under those idealized circumstances.
Thanks for doing the math! Good to have a concrete number.
They might be able to achieve parity with fiber on optical links between satellites -- basically the same technology that's used with fiber -- but of course atmospheric conditions won't support that sort of bandwidth at optical frequencies, so there would be no easy way to get it to/from the surface.
Let's just say there isn't enough "usable" spectrum, which partly has resulted in this licensing regime. Different frequency signals have different propagation characteristics and so only certain frequency ranges are amenable to specific use cases. For one, higher frequencies are more easily absorbed by materials, so you only get a few 100 MHz of frequencies suitable for long distance transmissions. As an example, even for home WiFi, you will note that 2.4GHz signals will travel farther in your home than 5GHz signals. With today's tech, usable frequencies are a scarce resource.
While I'd love to have google fiber... it sure seems to be taking a long time to roll out. Will it ever be in small rural towns? Will it beat satellites there?
There's a big jump between dialup and low end broadband. I can get by on low end broadband (which I think in rural areas may end up being LTS) but I can't get by on dialup.
A big part of the plan is scooping up backhaul traffic-- latency for long distance communication will be very fast with a few laser-linked satellites, compared to many (hundreds?) of hops linked by glass fiber (wherein the speed of light is significantly slower).
That and the billion+ potential customers currently without any access at all can probably provide a very healthy revenue stream, even if they don't pay much at all.
> There just isn't enough radio spectrum available.
That is why my bet is on Musk. With lots of low orbit transmitters each covering a small area you have "space division" like in a cellular network. I.e. less endpoints competing for the bandwidth to the same transmitter.
40 comments
[ 2.8 ms ] story [ 94.5 ms ] threadI for one am still hoping we'll eventually get quantum internet that goes through the earth rather than around it.
As an aside, the lower speed of light through fiber optic cable is one reason many stock trading platforms use line-of-sight microwave for their primary transmission. For HFT, those uS (and even nS) matter.
The lowest latency satellite systems, using LEO orbit satellites, are around 40ms roundtrip.
https://en.wikipedia.org/wiki/Anycast
It's that 15GB costs ~$100 and comes with an automatic next throttle on next month's bandwidth if I go over. Or, purchase additional GB for $10 each.
It's like having a Porsche with a thimble sized tank of gasoline.
> After exceeding either of the usage thresholds, which vary by service package, we may significantly slow and/or restrict your service, or certain uses of your service. Your full service will be restored when your data usage in the prior 30 days falls below 70% for both usage thresholds.
So, if you go over, then for the next 30 days you're throttled UNTIL your volume stays under 70% of your cap over a 30 day period... Like, how is it even legal to sell one thing, but punish for using more than 70% of it sometimes
Most providers sell consumer broadband access based on headline speed because that's what the fixed line providers do.
When buying "professional" satellite broadband; users can buy un-contented at much lower speeds. A Ford Transit with a dedicated lane perhaps...
"Between here and the satellite there is a very... large... number... of nanoseconds."
[1] http://www.geekwire.com/2015/photos-spacex-founder-elon-musk...
I recently saw an article about impulse engines using an arc-welding type approach where just electricity and some sort of metallic rod were needed. If these can be made small enough that could bring down the cost of station keeping dramatically, since presumably you could get a lot more lifetime thrust per kg of mass using that metal, then you would with propellants. (And the electricity can be harvested in orbit.)
I have Viasat Exede at my rural home that is unserved by other network products (no mobile signal, no CATV, no DSL).
Ping times are ~700ms and given my high reliance on SSH to do my work I was skeptical ... but it's not that bad.
I can't speak to gaming, etc., but for interactive login and things like skype/vonage it works OK.
Or local console login on a FreeBSD server at home.
Nothing special.
That, ontop of a whole host of other reasons, is why it's amazing. Absolutely love it, and have been using it every day for over a year now. Despite frequent internet drops (yay, comcast) and laptop closures, i only reconnect to the server when i reboot my laptop (maybe once a month, at best). It's neato!
edit: Note that my "server" work is my IDE (vim), and many panes of tmux, etcetc. Basically all my work. So it's used over 50h a week.
That being said, I hope one of these projects takes off. Most of the undeveloped world would benefit greatly from even just 5mb/s of internet.
This isn't really the goal of the project. SpaceX needs to create a com link with Mars. Also, it can cheaply laucnch sattelites so it will be able to connect the world via satelite uplink. The goal isn't to beat wired networks but provide redundancy and in some cases primary connections.
SpaceX is betting on itself achieving reusability which will lower costs from ~$60M to a marginal cost of something like ~$600K. As a byproduct of the cellphone wars nanosats have become insanely cheap due to commoditization and miniturization of hardware. Making this more feasible
The reason they are doing this is to establish communications between earth & mars. Even a 50mbps connection could bring huge swathes of the world online and create a race for infrastructure to compete.
So if we could get 1-5gbps connections terrestriallydue to some market penetration and competition, it would bringlots of people online. The goal of the project is lofty, but would provide a lot of positive things.
Downsides: * personal satelites and the proliferation of spying. * space debris could lock us out of orbit. * space debris/launch failures could cause issues. * something like radiation or an unexpected externality could make this non-feasible. * we could literally create skynet.
I am pumped on it though, and those reasons, along with the many others I have assuredly missed make me really really excited to see this happen. We need a secondary network anyway and terrestrial speeds won't improve without competition so I would roll the dice on skynet for that.
datenwolf did the math and answered it in greater detail than I.
Didn't Iridium try to do this a while back? Launch a constellation of LEO satellites to provide low latency data and cell communications? Hopefully the reduced launch and operation costs expected will give these current projects more success.
I am not very knowledgeable about things related to radio spectrums and so please clarify this for me.
What might the upper bound of radio broadcast data transmissions be if the entire radio spectrum could be allocated and given our current radio comms technology?
> When you say there isn't enough, is that there isn't physically enough spectrum, or is it just that there isn't enough licensable spectrum (or even enough available unlicensed spectrum)?
Both. Or rather physical spectrum is a superset of licencable spectrum and already the physical spectrum only has so many quantum numbers to encode information in.
We can pretty much rule out orbital momentum encoding for satellite transmissions; the effects the atmosphere has on polarization are just too strong, for orbital momentum to survive to a degree, that it remains usable for consumer grade equipment.
That leaves as encodable quantum numbers: - frequency - amplitude - polarization - momentum (i.e. direction of propagation)
The band usable for radio satellite communication, i.e. able to pass through the atmosphere is rather narrow (compared with the whole EM spectrum): About between 100MHz up to 20GHz. That leaves you with a total usable bandwidth of about 19.9GHz.
Multiply that with the symbol density reachable through AM (under good conditions with a low noise floor you can encode about 5 to 6 bits). Polarization gives another bit. Spatial multiplexing gives you log2(n steerable directions) extra bits (usually not more than 4 beams you can control). Divide that by two, because existing transceiver systems (except for some lab systems) can't do full duplex on a given channel.
If you multiply that up this gives you a theoretical limit of about 10TBit/s you can implement for a given satellite location (having several satellites in proximity means they have to share the spatial encoding bits).
Good answer by the way. I can second that what datenwolf said is pretty darn accurate, but doesn't account for a lot of the inteference that happens to reduce usability of the specturm.
You also get into beam widths so and antenna. So if you have 700 satellites up in the air given full spectrum that means they'd have a fixed limit of 10TBit/s. That can be divided between their users (but it won't be anything close to 10TBit/s).
So this will really be a factor of how big their beams are, how many people are within their beams, and how much frequency they can legally use. While this is really cool and possible, I doubt this will open up a whole new internet in it's first pass.
There is also the issue with, how are they going to get the signal to it's destination? They can't beam it all the way around the world for you, and from what I understand at least OneWeb isn't going to communicate from satellite to satellite. So they'll have to have a ton of ground stations to receive your link and then send it over fiber, or transmit up to GEO, then down again. OR they'll have to hop up and down between nodes to it's destination, in a similar fashion to TOR. I don't know how they plan to solve this or if SpaceX has a better solution.
They might be able to achieve parity with fiber on optical links between satellites -- basically the same technology that's used with fiber -- but of course atmospheric conditions won't support that sort of bandwidth at optical frequencies, so there would be no easy way to get it to/from the surface.
There's a big jump between dialup and low end broadband. I can get by on low end broadband (which I think in rural areas may end up being LTS) but I can't get by on dialup.
That and the billion+ potential customers currently without any access at all can probably provide a very healthy revenue stream, even if they don't pay much at all.
That is why my bet is on Musk. With lots of low orbit transmitters each covering a small area you have "space division" like in a cellular network. I.e. less endpoints competing for the bandwidth to the same transmitter.